Deep Amplicon Sequencing Reveals Culture-dependent Clonal Selection of Mycobacterium tuberculosis in Clinical Samples

Abstract

The commonly-used drug susceptibility testing (DST) relies on bacterial culture and faces shortcomings such as long turnaround time and clonal/subclonal selection biases. Here, we developed a targeted deep amplicon sequencing (DAS) method directly applied to clinical specimens. In this DAS panel, we examined 941 drug-resistant mutations (DRMs) associated with 20 anti-tuberculosis drugs with only 4 pg of initial DNA input, and reduced the clinical testing time from 20 days to 2 days. A prospective study was conducted using 115 clinical specimens, predominantly positive for the Xpert® Mycobacterium tuberculosis/rifampicin (Xpert MTB/RIF) assay, to evaluate DRM detection. DAS was performed on culture-free specimens, while culture-dependent isolates were used for phenotypic DST, DAS, and whole-genome sequencing (WGS). For in silico molecular DST, our result based on DAS panel revealed the similar accuracy to three published reports based on WGS. For 82 isolates, application of DAS using the resistance-determining mutation method showed better accuracy (93.03% vs. 92.16%), sensitivity (96.10% vs. 95.02%), and specificity (91.33% vs. 90.62%) than WGS using the Mykrobe software. Compared to culture-dependent WGS, culture-free DAS provides a full picture of sequence variation at the population level, exhibiting in detail the gain-and-loss variants caused by bacterial culture. Our study performs a systematic verification of the advantages of DAS in clinical applications and comprehensively illustrates the discrepancies in Mycobacterium tuberculosis before and after culture.

Keywords: Mycobacterium tuberculosis; Culture-dependent; Culture-free; Deep amplicon sequencing; Drug susceptibility testing.

Figure 1

Comparison of DST methods and overview of the DAS data analysis workflow A. Comparison of workflows for Xpert MTB/RIF, targeted DAS, WGS, and pDST assays. B. Overview of the DAS data analysis workflow. DST, drug susceptibility testing; DAS, deep amplicon sequencing; Xpert MTB/RIF, Xpert^®  Mycobacterium tuberculosis/rifampicin; DRM, drug-resistant mutation; MTB, Mycobacterium tuberculosis; MTBC, Mycobacterium tuberculosis complex; PATRIC, Pathosystems Resource Integration Center; mDST, molecular DST; WGS, whole-genome sequencing; pDST, phenotypic DST; TB, tuberculosis; MDR, multidrug-resistant; RF, random forest; RIF, rifampicin; pos, positive; neg, negative; cfDAS, culture-free DAS; QC, quality control; cdWGS, culture-dependent WGS; cdDAS, culture-dependent DAS; RDM, resistance-determining mutation.

Figure 2

Evaluation of detection capability of the DAS panel A. Number of variants in DRMs. B. Number of variants in all target regions. The DAS data from the serially diluted genomic DNA samples were compared to the WGS data from the initial DNA sample. Repeat1 and Repeat2 indicate the two replicates of the gradient-dilution assay.

Figure 3

Variant comparison in 83 clinical samples among cdDAS, cdWGS, and cfDAS data A. Allele frequency distribution in cdDAS, cdWGS, and cfDAS data (bandwidth = 0.01). B. Number of different loci between cfDAS and cdWGS data. C. Distribution of DRMs with different allele frequencies between cfDAS and cdWGS data. D. Similarity between cdDAS and cdWGS data based on target loci. E. Correlation of allele frequencies between cdDAS and cdWGS data. F. Similarity between cfDAS and cdWGS data based on target loci. G. Correlation of allele frequencies between cfDAS and cdWGS data. H. Box plot showing the differences in allele frequency between any two of cfDAS, cdDAS, and cdWGS datasets.

Figure 4

DRMs affecting mDST on clinical specimens before and after culture The alluvial plot generated by the easyalluvial package in R was used to visualize the connection for mutations. The mutations were clustered by sample, drug, frequency in cfDAS data, frequency in cdWGS data, gene, and mutation. Different colors stand for different drugs. EMB, ethambutol; RFB, rifabutin; FQ, fluoroquinolone; INH, isoniazid; ETO, ethionamide; AK, amikacin; KM, kanamycin; SM, streptomycin.

Figure 5

Allele frequency change between cfDAS and cdWGS data A. Allele frequency of DRMs in S001 (TBX1) increased in cdWGS data. B. Allele frequency of DRMs in S074 (TBX74) increased or decreased in cdWGS data. C. Allele frequency of DRMs in S077 (TBX77) decreased in cdWGS data. D. Allele frequency of gyrA:D94G in S295 (TBX295) increased together with the FQ treatment in cdWGS data. CM, compensatory mutation.